Statistical shape models

The shape of an object can be represented by a set of landmark points. The way that shape varies across a population can be modelled by first aligning shapes (sets of points) into a common reference frame, then applying Principal Component Analysis to learn the main modes of shape variation. Such models can be used to encode shapes and to help locate the landmarks in new images. This chapter introduces the mathematics behind such models and describes algorithms which can be used to locate the landmarks on new images.